How to utilize quadratic features for namespaces longer than one character

[alexcannan]

Description

I've been happily experimenting with Vowpal Wabbit recently and have become confused about the implementation of quadratic/cubic features. I have several namespaces inside my training that are all longer than one character. I've confirmed the data is syntactically correct, and can train all quadratic/cubic features with -q :: and -c :::. However, I'd like to pick exactly which feature combinations to use, say categories and text. I'm unable to find documentation on how to do this, and no delimiter I can think of works: -q categoriestext, -q categories*text -q categories^text, -q categories text. Is this behavior possible directly through the CLI?